[
https://issues.apache.org/jira/browse/BEAM-5779?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16657012#comment-16657012
]
Udi Meiri edited comment on BEAM-5779 at 10/19/18 4:16 PM:
-----------------------------------------------------------
The log from 6317 shows TestDataflowRunner attempting to cancel the job at
about 17:27:40 (expected since streaming jobs never end).
However, the job is still starting up and fails at 17:28:07 with the message:
Workflow failed. Causes: Step
"setup_resource_projects/apache-beam-testing/subscriptions/psit_subscription_input1a6def20-8a5b-4de7-a10b-609290f545b4__df_internale53817409a1b9ef015"
failed., Step
setup_resource_projects/apache-beam-testing/subscriptions/psit_subscription_input1a6def20-8a5b-4de7-a10b-609290f545b4__df_internale53817409a1b9ef015:
Set up of resource
projects/apache-beam-testing/subscriptions/psit_subscription_input1a6def20-8a5b-4de7-a10b-609290f545b4__df_internale53817409a1b9ef0
failed, Getting metadata for pubsub subscription
projects/apache-beam-testing/subscriptions/psit_subscription_input1a6def20-8a5b-4de7-a10b-609290f545b4
failed with error: Resource not found
(resource=psit_subscription_input1a6def20-8a5b-4de7-a10b-609290f545b4).
I believe increasing TEST_PIPELINE_DURATION_MS will fix the flakiness, which
affects how long we wait for the pipeline to complete.
I think I set it aggressively low at 30s, while streaming_wordcount_it_test
uses 3m.
was (Author: udim):
The log from 6317 shows TestDataflowRunner attempting to cancel the job at
about 17:27:40 (expected since streaming jobs never end).
However, the job is still starting up and fails at 17:28:07 with the message:
Workflow failed. Causes: Step
"setup_resource_projects/apache-beam-testing/subscriptions/psit_subscription_input1a6def20-8a5b-4de7-a10b-609290f545b4__df_internale53817409a1b9ef015"
failed., Step
setup_resource_projects/apache-beam-testing/subscriptions/psit_subscription_input1a6def20-8a5b-4de7-a10b-609290f545b4__df_internale53817409a1b9ef015:
Set up of resource
projects/apache-beam-testing/subscriptions/psit_subscription_input1a6def20-8a5b-4de7-a10b-609290f545b4__df_internale53817409a1b9ef0
failed, Getting metadata for pubsub subscription
projects/apache-beam-testing/subscriptions/psit_subscription_input1a6def20-8a5b-4de7-a10b-609290f545b4
failed with error: Resource not found
(resource=psit_subscription_input1a6def20-8a5b-4de7-a10b-609290f545b4).
I believe increasing TEST_PIPELINE_DURATION_MS, which affects how long we wait
for the pipeline to complete.
I think I set it aggressively low at 30s, while streaming_wordcount_it_test
uses 3m.
> Failure in beam_PostCommit_Python_Verify PubSubIntegrationTest on Dataflow
> --------------------------------------------------------------------------
>
> Key: BEAM-5779
> URL: https://issues.apache.org/jira/browse/BEAM-5779
> Project: Beam
> Issue Type: Bug
> Components: sdk-py-core
> Reporter: Kenneth Knowles
> Assignee: Udi Meiri
> Priority: Major
> Labels: flake
>
> https://builds.apache.org/job/beam_PostCommit_Python_Verify/6296/
> Path to failure:
> * (Jenkins) beam_PostCommit_Python_Verify
> * (gradle) :beam-sdks-python:postCommitITTests
> * (py module) test_streaming_data_only
> * (test class)
> apache_beam.io.gcp.pubsub_integration_test.PubSubIntegrationTest
> {code}
> AssertionError:
> Expected: (Test pipeline expected terminated in state: RUNNING and Expected 2
> messages.)
> but: Expected 2 messages. Got 0 messages. Diffs (item, count):
> Expected but not in actual: [('data002-seen', 1), ('data001-seen', 1)]
> Unexpected: []
> {code}
> Looks like a legitimate failure. It is flakey as it went green later.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
